As user requirements for bandwidth are growing, a conventional copper-wire broadband access system is increasingly facing a bandwidth bottleneck. In addition, as fiber optic communications technologies with a huge bandwidth capacity become increasingly mature and application costs decrease year by year, a fiber access network becomes a promising candidate for a next-generation broadband access network. In particular, a passive optical network (PON) system is most competitive. FIG. 1 shows a network architecture of a PON system. The PON system usually includes an optical line terminal (OLT) located at a central office, an optical distribution network (ODN) used for branching/coupling or multiplexing/demultiplexing, and several or multiple optical network units (ONUs).
In the PON system, as a physical implementation body of a transceiver system, an optical module is an essential device. As an apparatus for transmitting an information carrier and a communication lightwave, a laser is even more important. In many application occasions of the PON field, we need to keep an emission wavelength of the laser at a specific value to ensure compliance with an optical communications standard or technical specifications related to physical transfer. Therefore, in many application scenarios, the laser of the optical module usually includes a semiconductor-based thermoelectric (TEC) or a heating membrane to adjust a wavelength of the laser, and a specific wavelength monitoring (or wavelength alignment) optical path is required to implement feedback adjustment. In addition, an optical filter-based laser wavelength locking technology is currently the most popular and most efficient technology.
FIG. 2 shows a schematic diagram of a wavelength alignment between a laser and a filter. In the optical filter-based laser wavelength locking technology, the filter is used to select a laser wavelength mode. Because the filter has different transmissivity (or reflectivity) for laser light of different wavelengths, when the laser wavelength is shifted, an optical power changes after the laser light passes through the filter. A shift of the laser wavelength can be monitored in this way. As shown in FIG. 2, FIG. 2 is a typical multi-channel comb filtering curve. W1, W2, and W3 are laser wavelengths, and A, B, and C are positions in filtering bevel edges of a filter that are respectively corresponding to the laser wavelengths. A wavelength change of a filter is most directly reflected in a change in a position in a filtering bevel edge of a filter, and this results in a change in transmissivity (or reflectivity) of the filter light. Wavelength locking is to ensure that a position that is in a filter wave and that is corresponding to a wavelength is at a peak of a filtering curve, that is, corresponding to a position where maximum transmissivity is available.
FIG. 3 shows a transmitter optical sub-assembly (TOSA) packaging structure that is based on a common wavelength locking manner. A laser (designated as DFB for distributed feedback laser), a collimation lens, an isolator, an optical filter (designated as Etalon), and a focusing lens are connected in a line. Herein, because an interferometric Etalon requires injection of collimated light, the collimation lens needs to be added between the DFB and the Etalon. In addition, to prevent reflected light from entering the DFB device to affect performance, the isolator is added between the collimation lens and the Etalon. In this way, light reflected by the Etalon and reflected by a loop does not enter the DFB. Because the Etalon is required to have a filter model selection function, a reflection function is required. In addition, reflection from a fiber loop is also unavoidable. In this solution, to prevent impact from reflection, the isolator is added to a hermetic chamber. However, this optical structure that is tiled in a straight line has a relatively large overall size, and consequently, transistor outline (TO) packaging cannot be implemented. Therefore, the packaging can only be implemented using X Mini Dimension (XMD) (where “X” is the Roman numeral for 10) that is more expensive.
FIG. 4 shows a TO optical wavelength locking manner that is based on a filter tilt mode. In this solution, the isolator is taken out of a hermetic chamber of a TOSA and is disposed outside the hermetic chamber. In this case, a packaging volume of a hermetic part can be reduced such that the original XMD packaging can be changed to low-cost TO can (TO-Can) packaging. Certainly, in this case, an optical reflection problem caused after the isolator is taken out needs to be resolved. The reflection from the fiber loop can still be eliminated by the isolator that is disposed outside the hermetic chamber. However, reflection from an internal end face of the Etalon still exists. In this solution, a method used to resolve the reflection from the Etalon is to make an incident surface of the Etalon tilt for a specific angle such that reflected light is shifted by a specific angle and is prevented from entering the DFB to cause impact, as shown in FIG. 4.
When the manner of tilting the Etalon for an angle, as shown in FIG. 4, is used to prevent reflection, some unavoidable costs need to be paid, mainly including:
(1) Because the Etalon is a resonant device, a peak insertion loss increases if the Etalon is tilted for a specific angle. As shown in FIG. 4, when the angle is 4 degrees, an extra insertion loss already reaches 1.5 decibel (dB).
(2) A tilt angle also causes broadening on a full width at half maximum. When the angle is 4 degrees, filtering performance has deteriorated by 20 gigabits per second (Gbit/s), and a filtering effect is greatly reduced.
Consequently, an increased insertion loss causes a decreased system power budget, and the broadening on the full width at half maximum directly causes decreased accuracy of optical wavelength locking of a signal. In addition, a greater peak insertion loss is caused when the full width at half maximum of a filtering curve is compressed by increasing reflectivity of front and rear end faces (refer to FIG. 5 front and rear aspheric surfaces) of the collimation lens. Therefore, this disadvantage is unavoidable.